Open belforte opened 7 years ago
reference: https://hypernews.cern.ch/HyperNews/CMS/get/computing-tools/2450/1.html
Also I do not like that gfal-copy is called with decreasing timeouts, but that's a different story. The management of the timeout here is not friendly, but maybe we can simply put better messages and a bit of documentation to push users into tuning the amount of files to copy in each command.
Something maybe due to "thread confusion" ? See following[1] fragment of crab.log, some copy command have the correct
-t <timeout>
while all of a sudden one queue command has--srm-timeout <timeout>
which is the one for lcg-copy. This makes no sense from the code, there is no indication that the option could be different from one file to the next one. https://github.com/dmwm/CRABClient/blob/master/src/python/CRABClient/Commands/remote_copy.py#L147-L150The hypothesis of thread confusion is also because earlier in the log I find [2]
[1]
[2]